Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220202102147.326672-1-linux@rasmusvillemoes.dk>
In-Reply-To: <20220202102147.326672-1-linux@rasmusvillemoes.dk>
From:   Nick Desaulniers <ndesaulniers@google.com>
Date:   Wed, 2 Feb 2022 13:04:52 -0800
Message-ID: <CAKwvOd=h7NLaqtRwA3FTsq_T-=_0XKL0GRA49CRg5F19m=7xPg@mail.gmail.com>
Subject: Re: [PATCH] include: drop pointless __compiler_offsetof indirection
To:     Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc:     Miguel Ojeda <ojeda@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Nathan Chancellor <nathan@kernel.org>,
        Kees Cook <keescook@chromium.org>,
        linux-kernel@vger.kernel.org, llvm@lists.linux.dev
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Feb 2, 2022 at 2:21 AM Rasmus Villemoes
<linux@rasmusvillemoes.dk> wrote:
>
> (1) compiler_types.h is unconditionally included via an -include
> flag (see scripts/Makefile.lib), and it defines __compiler_offsetof
> unconditionally. So testing for definedness of __compiler_offsetof is
> mostly pointless.
>
> (2) Every relevant compiler provides __builtin_offsetof (even sparse
> has had that for 14 years), and if for whatever reason one would end
> up picking up the poor man's fallback definition (C file compiler with
> completely custom CFLAGS?), newer clang versions won't treat the
> result as an Integer Constant Expression, so if used in place where
> such is required (static initializer or static_assert), one would
> get errors like
>
> t.c:11:16: error: static_assert expression is not an integral constant expression
> t.c:11:16: note: cast that performs the conversions of a reinterpret_cast is not allowed in a constant expression
> t.c:4:33: note: expanded from macro 'offsetof'
> #define offsetof(TYPE, MEMBER)  ((size_t)&((TYPE *)0)->MEMBER)
>
> So just define offsetof unconditionally and directly in terms of
> __builtin_offsetof.
>
> Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
> ---
>  include/linux/compiler_types.h | 2 --
>  include/linux/stddef.h         | 6 +-----
>  2 files changed, 1 insertion(+), 7 deletions(-)
>
> diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
> index 3c1795fdb568..83ee7f7ada5d 100644
> --- a/include/linux/compiler_types.h
> +++ b/include/linux/compiler_types.h
> @@ -137,8 +137,6 @@ struct ftrace_likely_data {
>   */
>  #define __naked                        __attribute__((__naked__)) notrace
>
> -#define __compiler_offsetof(a, b)      __builtin_offsetof(a, b)
> -
>  /*
>   * Prefer gnu_inline, so that extern inline functions do not emit an
>   * externally visible function. This makes extern inline behave as per gnu89
> diff --git a/include/linux/stddef.h b/include/linux/stddef.h
> index ca507bd5f808..929d67710cc5 100644
> --- a/include/linux/stddef.h
> +++ b/include/linux/stddef.h
> @@ -13,11 +13,7 @@ enum {
>  };
>
>  #undef offsetof
> -#ifdef __compiler_offsetof
> -#define offsetof(TYPE, MEMBER) __compiler_offsetof(TYPE, MEMBER)
> -#else
> -#define offsetof(TYPE, MEMBER) ((size_t)&((TYPE *)0)->MEMBER)

RIP `((size_t)&((TYPE *)0)->MEMBER)`; you were beautiful in a way only
a mother could love.
Acked-by: Nick Desaulniers <ndesaulniers@google.com>

> -#endif
> +#define offsetof(TYPE, MEMBER) __builtin_offsetof(TYPE, MEMBER)
>
>  /**
>   * sizeof_field() - Report the size of a struct field in bytes
> --
> 2.31.1
>


-- 
Thanks,
~Nick Desaulniers